1. Field of the Invention
The present invention relates to technology that can classify data of images respectively including different subjects having similar faces, such as parent and child or siblings, into different groups with high accuracy.
2. Related Art
Recently, memory storage sizes are increasing, and digital cameras have become capable of storing data of an extremely large number of images.
In such cases, with the increase in the number of stored images, operations of searching for data of a desired image and organizing data of images that have been searched for is becoming difficult.
In order to reduce such operational difficulty, there is known a technique that extracts a region of a face (hereinafter, referred to as “face region”) included in an image by way of face detection technology, associates data of the image from which a face is detected with a specific group, and thereby classifies data of a plurality of images into groups of respective subjects.
However, since data of different images from which identical or similar face regions are detected is classified into the same group, there may be cases in which a group assigned to a predetermined person happens to include image data of a different person having a similar face to the person.
For example, consideration is given to a case of classifying data of a first image including a predetermined person as a subject and data of a second image including a different person as a subject, who has a parent-child or sibling relationship with the predetermined person.
In this case, if both the first and second images are photographed at the same time, e.g., at the present time, since the two subjects are different in age, it is possible to classify data thereof into different groups. This is because, when comparing a first face region detected from the first image and a second face region detected from the second image, there are differences in the features due to age difference therebetween, more specifically, horizontal to vertical ratio of the faces, height of eyes on the faces, size of eyes, number of wrinkles, or recession of hair. Therefore, it is possible to classify data of the first and second images into different groups according to difference in such features.
However, in a case in which the predetermined person is, for example, a parent, the subject included in the first image is the parent at the age of 4, the different person is a child, and the subject included in the second image is the child at the age of 4, then it becomes difficult to classify data of such images into different groups. This is because the two subjects are the same in age, and there are no differences in features in terms of age difference as described above. Furthermore, faces of a parent and a child at the same age are most likely very similar. Therefore, in such a case, the respective data of the first and second images cannot be separated, and the first and second images may be classified into the same group of either the parent or the child.
For this reason, it is desired to realize a method capable of classifying data of images respectively including different subjects having similar faces such as parent and child or siblings into different groups at high accuracy.